Consistency and diversity neural network multi-view multi-label learning

作者：

Highlights：

• A MVML feedforward neural network model is constructed, which is different from the traditional neural network model.

• This model effectively considers the issues of view consistency, view diversity, and label correlations in MVML learning. In addition, a similar ensemble learning method is used to predict the final model.

• The large number of empirical results of CDMM on the benchmark data set proves that it has certain advantages compared with some related and competitive methods.

摘要

•A MVML feedforward neural network model is constructed, which is different from the traditional neural network model.•This model effectively considers the issues of view consistency, view diversity, and label correlations in MVML learning. In addition, a similar ensemble learning method is used to predict the final model.•The large number of empirical results of CDMM on the benchmark data set proves that it has certain advantages compared with some related and competitive methods.

论文关键词：Multi-view learning,Multi-label learning,Consistency,Diversity,Neural network

论文评审过程：Received 4 September 2020, Revised 20 December 2020, Accepted 2 February 2021, Available online 10 February 2021, Version of Record 15 February 2021.

论文官网地址：https://doi.org/10.1016/j.knosys.2021.106841